Tactile-kinesthetic-proprioceptive (TKP) input used to facilitate speech motor control is considered an active ingredient within speech motor interventions. Objective metrics identifying skill level differences across speech-language pathologists (S-LP) providing TKP cues are crucial for monitoring treatment delivery fidelity. The study examined three kinematic measures indicating accuracy and consistency of TKP inputs by 3 S-LPs with varying experience levels (S-LP 1: novice; S-LP 2 and S-LP 3: advanced). Confidence interval measures were used to compare the accuracy of jaw movement amplitudes of the vowel /a/ made by a model participant versus S-LPs giving the TKP input. Generalised Orthogonal Procrustes Analysis (GPA) and cyclic Spatial Temporal Index (cSTI) were used to determine movement consistency. Results revealed passive jaw excursions induced by S-LP 2 and 3 to be not statistically significant from the model participant’s active jaw movements. cSTI values decreased with advanced level of experience (19.28, 12.14, and 9.33 for S-LP 1, S-LP 2, and S-LP 3, respectively). GPA analyses revealed a similar pattern for S-LPs with more experience demonstrating lower mean RMS values (0.22, 0.03, and 0.11 for S-LP 1, S-LP 2, and S-LP 3, respectively). Findings suggest kinematic measures adapted from the motor control literature can be applied to assess S-LP skill differences in providing TKP cues.
The use of tactile-kinesthetic-proprioceptive (TKP) inputs has an established presence in remediating speech sound disorders (SSDs) within many contemporary speech motor interventions (e.g., [
The effectiveness of TKP cues in improving accuracy of words and phrases, facilitating movement generalization, and establishing, refining, and integrating normalized speech movement patterns within several articulatory subsystems (e.g., mandibular, labial-facial, and lingual) has been demonstrated in a number of behavioral studies (e.g., [
Given the importance of TKP inputs in sensory-motor mapping and speech movement stability (see [
However, some fidelity assessment procedures have been described and utilized by researchers in communication disorders [
There is consensus among a number of theoretical perspectives that learning complex skills is a process occurring over several time scales [
The current study sought to evaluate whether three objective kinematic measures relating to the accuracy and consistency of S-LP hand movements and the way these movements are able to match resulting passive orofacial movements to active (natural) movement ranges can be used to distinguish experience levels of S-LPs providing those TKP cues. For the purpose of this pilot study, a model participant without speech problems was deemed to provide a more reliable context for testing the usefulness in assessing clinicians’ skills. We hypothesize that, with more specialized training and years of experience with speech motor disorders, a S-LP should (a) demonstrate increased accuracy when predicting and cuing client jaw ranges for a given vowel (relative to the 95% confidence intervals (CIs) of the client’s mean range of active jaw movements), (b) demonstrate increased consistency in finger-hand movements when providing TKP cues as determined by Generalised Orthogonal Procrustes Analysis (GPA), and (c) demonstrate increased consistency in induced (passive) orofacial movements using a cyclic Spatial Temporal Index (cSTI) measure. These measures will be explained in more detail in the method section.
A 23-year-old female with no self-reported speech, language, hearing, or neurological difficulties served as the model participant-client in this study. Three female S-LP participants, with different levels of training and experience (S-LP 1: novice; S-LP 2 and S-LP 3 advanced), provided TKP cues to the model participant’s orofacial structures. S-LP 1 had recently graduated from an accredited S-LP program with eight months experience providing S-LP services to children with speech sound disorders. S-LP 2 and S-LP 3 had more than 15 years of experience and advanced training with speech motor interventions with both children and adults. The study was approved by the University of Toronto’s Health Sciences Research Ethics Board, and all participants provided a written informed consent prior to participation.
All kinematic and time-aligned acoustic data were collected with the Electromagnetic Articulograph 501 system (EMA AG501) [
Sensors were attached onto the model participant’s face at the following anatomical locations: mandible midline, nose bridge, behind the left and right ear mastoid, and two sensors 1.5 cm symmetrically on either side of the philtrum of the upper lip. Furthermore, six sensors were placed on the hand providing the TKP inputs in the following manner: one each on the thumb, index finger, and middle finger (placed on the nail plate) and three reference sensors on the dorsal part of the hand.
Speech stimuli were vowels /a, i, u/ produced in isolation and in a sequence combined with the consonant /t/ as in /ta, ti, ta, tu/ in succession. Movement data were recorded under two conditions: (a)
Three kinematic measures were used to quantify accuracy and consistency of TKP cues provided by an S-LP. All of these measures have been modified and adapted from measures previously reported in the speech and limb motor control studies [
Measuring accuracy of jaw movements has been deemed of crucial importance in enhancing the critical role of jaw control in speech production [
S-LP finger placement and orientation for TKP inputs related to vowel /a/ (a), vowel /i/ (b), and vowel /u/ (c).
Descriptive statistics mean (standard deviation) and number of attempts (
S-LP | Participant | Unpaired |
---|---|---|
S-LP 1: 6.46 (1.12), |
11.88 (2.66), |
|
S-LP 2: 10.33 (1.30); |
9.18 (0.78), |
|
S-LP 3: 4.42 (0.71); |
5.39 (2.05), |
|
Two measures were utilized to quantify the consistency of TKP inputs provided by an S-LP: the Generalised Orthogonal Procrustes Analysis (GPA) and the cyclic Spatial Temporal Index (cSTI). For vowel /i/, the thumb and index fingers were placed above the upper lip at the intersection between levator anguli oris and the skin locations covering the location of zygomatic major muscles with slight pressure applied backwards towards the model participant [
Generalised Orthogonal Procrustes Analysis [
The entire movement trajectory data from the S-LP's thumb (relative to the reference sensor on the participant’s nose bridge) over successive trials of stimuli “ta-ti-ta-tu” was used for the GPA analysis. To extract movement data of the S-LP's thumb relative to the reference sensor, the latter was subtracted from the trajectory of the S-LP’s thumb in 3 dimensions (X = front/back, Y = left/right, and Z = up/down). Before applying GPA, each movement trajectory was time normalized to 1000 points using FFT (FFT is an algorithm that completes the discrete Fourier transform of a sample of points. During FFT interpolation, a sequence of values across time is resolved into the frequency domain using the FFT algorithm. These frequencies are then converted back into the time domain using inverse Fourier transform, and however, the sampling frequency is kept different from that initially used. In this manner, we are able to change the sampling frequency (hence number of points) in a time-domain signal, while preserving the underlying frequency values of the signal itself) interpolation separately for each of the three dimensions, across 10 repetitions of /ta-ti-ta-tu/. A centroid or mean value was calculated for each movement trajectory and for each dimension (X, Y, and Z). The movement trajectories were then linearly shifted such that the centroids are aligned with the origin (noise bridge sensor coil). The thumb movement data for each of the 10 repetitions of /ta-ti-ta-tu/ was then represented as a 1000∗3 matrix, with each of the rows representing 3D positions at a particular time and each column representing the position value along one of the three Cartesian coordinates across time. For convenience in describing the GPA method, we will follow the convention of referring to each of the 10 1000∗3 matrices as Si (where i represents the ith repetition). During the GPA process, we attempt to align repeated 3D movements. In our case, we have 10 repetitions. Each repetition contains a 1000 points of 3-dimensional data (hence a 1000∗3 matrix). Si is a convention to represent one such matrix. S1 would represent the first matrix representing the first repetition, S3 would represent the third matrix representing the third repetition, and so on.
For the GPA analysis, one movement trajectory of the 10 repetitions of /ta-ti-ta-tu/ was selected as a common consensus trajectory or exemplar (M), and the remaining nine movement trajectories were rotated to align with this exemplar. Matrix rotation is performed by finding an orthogonal matrix Qi for each movement trajectory Si which minimizes the value of ||Si∗Qi-M||. The solution for finding such an orthogonal rotation matrix Qi has been published elsewhere [
However, GPA assesses consistency in the shape of the S-LP's finger space movement paths, and we need a measure to capture end product consistency of a S-LP's finger-hand movement trajectories (i.e., the resulting changes in the orofacial structures, for example, skin stretch or deformation). In the current study, cyclic STI or cSTI [
Kinematic consistency of passively induced upper lip movements was indexed using the cSTI, with movement cycles operationally defined as peak-to-peak or valley-to-valley trajectory cycles related to the Euclidean distance between the two sensor coils placed 1.5 cm symmetrically on either side of the model participant’s philtrum on the upper lip. For cSTI, cycles were limited to displacement records for /ti/ for which the S-LP was inducing lip retraction gestures in the model participant. Lip gestures corresponding to /tu/ (for lip rounding) were not used in the cSTI analysis due to motion artifacts caused by lip muscle protrusion and bunching (resulting in sensor coil rotation) from the S-LP’s fingers pulling the lips closer. The lip retraction displacement records for /ti/ gestures were time (to 1000 points) and amplitude (
Descriptive statistics (Mean and Standard Deviations (S.D.)) for passive jaw excursions (in mm) induced by S-LP 1, 2 and 3 compared to active jaw movements made by the model participant are provided in Table
Jaw accuracy data (Table
Accuracy data across 3 S-LPs. Confidence Interval (95%) of jaw movement range for actively produced vowel /a/ by the model participant compared to induced movement ranges by the 3 S-LPs. Each cross represents an S-LP attempt, approximately 10 attempts per S-LP.
In Figure
GPA analysis as applied to S-LP's thumb finger movement trajectories over successive trials of stimuli “ta-ti-ta-tu”. Since the movement paths are rotated, the trajectories after GPA do not correspond to the original movements in the Cartesian axis. Hence, by convention, we use the principal axes of the GPA consensus path in the plots. Data from (a) S-LP 1, (b) S-LP 2, and (c) S-LP 3.
In Figure
Depicts cSTI values derived from the amplitude- and time-normalized displacement records (for /ti/- lip retraction induced by S-LP) from the two sensor coils placed 1.5 cm symmetrically on either side of the model participant’s philtrum on the upper lip (note: 100 points from the onset and offset were removed from analysis due to high frequency artifacts arising from windowing). Data from (a) S-LP 1, (b) S-LP 2, and (c) S-LP 3, respectively.
The purpose of the study was to identify objective measures to distinguish SLP’s skill levels in providing TKP cues. We investigated whether three kinematic measures relating to the accuracy and consistency of S-LP hand movements and the resulting model participant’s passive orofacial movements would distinguish experience levels of S-LPs providing TKP cues. Overall, we found that, in the task of accurately estimating a client jaw movement range, changes in consistency of an S-LPs own finger-hand movements when providing TKP cues and of the induced (passive) orofacial movements in a model participant varied as a function of the S-LP’s training and experience. These findings suggest that kinematic measures adapted from the speech and limb motor control literature can be successfully applied to quantify S-LP skill levels in providing TKP cues. The broader clinical implications of these findings will be discussed next.
The act of speaking requires coordination of the speech subsystems (respiration, phonation, and articulation) within time and space. Past literature has shown that the multiple degrees of freedom of movement in speech production are constrained through organized functional synergies, allowing for variability in movement patterns [
Fundamental to intervention for clients with SSDs are skills requiring attunement to task-specific visual information while observing a client’s ability to execute a speech motor movement, assessing where difficulties are occurring, and planning intervention by selecting speech motor targets using TKP inputs [
More specifically, for S-LP training, the S-LP needs to learn how to perform the motor act underlying the TKP cues, that is, learn to move his/her hand/fingers to a specific spatial location (i.e., client’s jaw/face), timed with the movement within the intended range [
Contemporary motor control literature considers the consistency with which behaviors are repeatedly executed to be one of the critical elements of skilled motor expertise [
The GPA is said to reflect the precision or computational difficulty with which the motor system executes the desired movement paths. The higher mean RMS residual values in S-LP 1 (least experienced) may imply a greater difficulty in planning and executing complex thumb movement paths as required for TKP inputs in a therapeutic context. The data from the more experienced S-LPs demonstrated that they executed their finger/hand movements with greater precision, indicated by the lower RMS values. In a study by [
From a clinical standpoint, consistency of induced movements in target structures such as the lips and jaw (i.e., end product of finger-hand cueing movement trajectories) is as important as consistency of S-LP hand movements. Experimental data suggests that somatosensory signals arising from cutaneous afferents in the facial skin (in the absence of muscle receptors in the perioral structures) play a crucial role in speech motor learning and adaptation [
The importance of establishing treatment fidelity to determine treatment efficacy has been clearly established [
Verrel et al. [
The study has several major limitations. External generalizability of these results is severely limited by the small sample of participants. A demonstration of these measures and outcomes with a larger group of participants is imperative. Furthermore, using a healthy model’s actual facial movements to judge appropriate TPK cueing, although appropriate for the context of this pilot study, does not allow for generalization to clients with impaired speech production skills. In the field of speech-language, pathology assessment of functional speech outcomes is critical. In therapy where these cues are typically used, such measures are indeed critical for assessing progress in a client’s speech production abilities. However, this information was not gathered in the present study as the focus was specifically on the accuracy and consistency measures obtained from the S-LP’s hand movements and S-LP-induced passive jaw and lip movements on a model participant’s face. This individual was not actually speaking during measurements of these variables (other than to establish the jaw movement range), again because our focus for the present study was not on the acoustic outcomes but on the delivery of tactile cues. This is why, for the purpose of this pilot study, we used an adult model participant without speech, language, hearing, or neurological issues so it would be possible to access the specific features of tactile cue delivery without possible confounding issues presented by a motor speech disorder. This presented a more reliable context for testing the feasibility of assessing clinicians’ skills. In a healthy adult participant with perfectly intact and “normal” functioning sensory-motor systems, we do not expect any significant changes in speech acoustics as a function of accuracy and consistency of TKP cues, especially with a limited number of trials (10 repetitions of target items). However, in the context of a developing sensory-motor system in a child or an impaired speech system in either adults or children (e.g., subsequent to brain injury), the auditory-to-speech motor mapping required for the achieving accurate speech output may be more susceptible to inaccurate and inconsistent TKP input by an S-LP. The findings from the current study suggest that kinematic measures adapted from the motor control literature can indeed be applied to assess S-LP skill differences in providing TKP cues. Utilizing these objective measures to capture S-LP skill levels, future studies must be conducted on populations with speech disorders to establish the impact of S-LP skill levels on functional speech outcomes.
Lastly, the cross-sectional design used provides no indication of the longitudinal development of motor skill in clinicians and limits causal inference. Future studies should track S-LPs’ motor skill levels as a function of training in a longitudinal study using a population with specific speech disorders and possibly within a therapeutic context. This may one day allow us to track S-LPs’ motor skill levels objectively and set fidelity standards for providing specialized speech treatments.
Overall, we found that accuracy in estimating and subsequently implementing a client’s jaw movement range, consistency in finger-hand movements when providing TKP cues, and the consistency of the resulting-induced (passive) orofacial movements in a model participant varied as a function of the S-LP’s training and experience. These findings suggest that kinematic measures adapted from the speech and limb motor control literature can be successfully applied to quantify S-LP skill levels in providing TKP cues. These preliminary findings are a first step for further explorations of the feasibility of these measures in evaluating S-LP functional skill levels during training and for determining if a clinician is ready to treat patients without supervision.
The data used to support the findings of this study have not been made available because they are restricted by the Health Sciences Research Ethics Board (University of Toronto) in order to protect patient privacy.
The authors declare that they have no conflicts of interest.